Quantification of Hepatitis Severity Degrees by Fuzzy Classifiers FCM and ANFIS
Anyta MUKAWA LUKENZU
Department of Mathematics, Statistics and Computer Science, Faculty of Science and Technology, National Pedagogical University, Kinshasa, DRC.
Jonathan OPFOINTSHI ENGOMBANGI
Higher Institute of Medical Techniques of Bandundu (ISTM BDD), Democratic Republic of Congo.
Camile LIKOTELO BINENE *
Department of Mathematics, Statistics and Computer Science, Faculty of Science and Technology, National Pedagogical University, Kinshasa, DRC.
Emilien Loranu Londjiringa
Exact Sciences Section, Department of Mathematics and Physics, ISP BUNIA, Ituri, DRC.
Grâce NKWESE MAZONI
Department of Mathematics, Statistics and Computer Science, Faculty of Science and Technology, National Pedagogical University, Kinshasa, DRC.
*Author to whom correspondence should be addressed.
Abstract
This study is not the first to address the diagnosis of hepatitis cases in patients.
We have analyzed several works of diagnosis of liver tumors (hepatitis) by laboratories using imaging but the most similar to our research is the one entitled: Integration of ontologies in the parallel classification of medical data for the diagnosis of hepatitis lesions, where we found that the laboratory results always remained unclear or vague when talking about hepatitis is in the early stage or in the advanced stage, without specifying whether the hepatitis is in the advanced stage with such or such degree.
In this article, we present a computer tool in artificial intelligence, based on the fuzzy classifiers FCM and ANFIS, capable of automatically determining the severity degrees of hepatitis in patients with two or more types at the same time, in order to minimize errors related to the fatigue of expert doctors faced with data complexity.
The aim of this article is to create a practical computer tool for determining the degree of hepatitis in patients, instead of simply saying that hepatitis is in the early or advanced stage and not to create new theories about FCM and ANFIS.
To achieve this, we started from a data set with only the inputs of 106 infected people, registered in the laboratories of H.JEHOVAH and H.G of de Kinkole, in the city province of Kinshasa (DRC) to perform the segmentation with the FCM algorithm. And the optimization constraint led us to have the sum of degrees of contamination equal to 1 or 100%.
Among the clinics of infected patients, we selected the following: temperature, age, weight, presence of fever, fatigue, and appetite. The order of the hepatitis types used in this article is as defined by (Tedros Adhanom, 2021), with the two blocks below, according to the WHO: B, C, and D as chronic hepatitis, and A, E as acute hepatitis.
Faced with the new observations, we used the outputs provided by the FCM segmentation algorithm to predict their severity degrees (reflecting reality) of the different types of hepatitis with the Neuro-fuzzy ANFIS algorithm based on fuzzy inference of Sugeno I and Mamdani.
Regarding the implementation language, we used MATLAB for segmentation and construction of the prediction model.
Among the major results, we have:
- The FCM was able to quantify the degree of belonging to the different types of hepatitis, at the early or advanced stage, which (Verma & Dubois, 2021) was not able to achieve.
- The FCM revealed the hidden types of hepatitis beyond the types of hepatitis recorded in the laboratories of H.JEHOVAH and H.G of de Kinkole (DRC).
- Hidden types of hepatitis are considered here as causes of resistance of hepatitis viruses despite their treatment.
Keywords: Severity degree, membership function, FCM, ANFIS, segmentation, prediction, output, linguistic variable, linguistic values